Text Generation WebUI

This project is now called TextGen, and it has moved from being a browser interface for local models to being a desktop application for them. Download a portable build, unzip it and open the window: the repository describes builds for Linux, Windows and macOS with acceleration for each major graphics vendor and a processor-only option, all dependencies included, and it says plainly that there is no telemetry and nothing reaches out to the network. Chat with instruction-tuned models or with a character, attach images to a message for vision models, attach documents to discuss their contents, edit messages and branch conversations; there is a notebook tab for generation outside a chat turn.

It suits people who want to switch between backends and models in one place without restarting, and it suits anyone who wants the OpenAI- or Anthropic-shaped API running locally as a drop-in replacement, including tool-calling and MCP servers. It also suits fine-tuning a small adapter and generating images from the same window. It suits less well anyone who wants a small, quiet dependency: the whole point is breadth, which is why the project ships several engines and a full install path beside the portable download, and why the licence it carries is the one that reaches into network use rather than the most permissive one.

Not verified: nothing was downloaded or run, so the portable builds, the backend switching and the API compatibility were not tested. The feature list, the claim of zero telemetry and the statement that models are detected automatically are the project's own description in its repository. The repository interface reported about 47,728 stars, 5,984 forks, the AGPL-3.0 licence and a push on 17 August 2026; the older name and the older browser-based form appear to have been retired, but no announcement of a rename was read, so whether the old address still resolves or an old install still upgrades was not checked. The documentation wiki was not read in full.